Fix creating schema from model using ConstrainedStr with regex str as dict key

[matejetz]

Change Summary

Currently, schema creation fails if a model uses a ConstrainedStr model as keys in a dict:

class CustomStr(ConstrainedStr):
    regex = '^text$'

class Model(BaseModel):
    a: Dict[CustomStr, Any]

json_schema = Model.schema()
# AttributeError: 'str' object has no attribute 'pattern'

ConstrainedStr regex field is typed as Optional[Union[str, Pattern[str]], when accessing pattern generating the schema for patternProperty, regex.pattern is accessed.

This PR contains the following changes:

• Use ConstrainedStr._get_pattern function in schema.py when accessing the pattern
• Add a test case for the described scenario

Checklist

• Unit tests for the changes exist
• Tests pass on CI and coverage remains at 100%
• Documentation reflects the changes where applicable
• changes/<pull request or issue id>-<github username>.md file added describing change
  (see changes/README.md for details)
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

Selected Reviewer: @samuelcolvin

[matejetz]

please review :)

[samuelcolvin]

We're trying to absolutely minimise the footprint in changes to v1, hence my suggestions.

Please update.

[matejetz]

updated towards minimal footprint, please review

[samuelcolvin]

otherwise LGTM.

@dmontagu please review and merge if you're happy.

[dmontagu]

LGTM. I don't like using the "private" ConstrainedStr._get_pattern in a separate module, but I think it's better than duplicating the (admittedly fairly simple) logic (since if it changed it would be easy to miss that it needed an update here), and I don't think it makes sense to make the method public.

If necessary, we can be more thoughtful about the method naming, etc., in v2 see note below, nothing should need changing in v2

[dmontagu]

Yep, it looks like the code in v2 has been refactored so this particular failure mode can't even happen. So I don't think there's anything to do in terms of implementing a similar fix in v2.